While not elbow deep in the flesh, blood, rubber and metal of surgical robotics, engineering researchers at the University of Florida are playing a key role in advancing safety and precision in robot-assisted surgery, or RAS.
Professor Christophe Bobda, Ph.D., and Ph.D. student Antonio Hendricks are analyzing vast datasets from robot-assisted procedures, processing them using artificial intelligence and finding patterns and processes that can be improved.
“We’re analyzing vast sums of data recorded by the robotic surgical system, such as stereoscopic video and kinematic data, via machine learning,” Hendricks said. “These processes are used to conduct autonomous surgical skill assessments and provide insights into interactions with respect to tissue handling and the final product.
They are gathering information using imaging and segmentation to know where everything is – “where the organs are, where the needles are, where the tools are,” said Bobda, the associate chair for academics for Electrical and Computer Engineering (ECE).
The goal: robot autonomy in the operating room in 10 years with easier surgical recoveries. To be clear, that does not mean operating rooms will be devoid of human surgical personnel. There will still be a surgeon controlling the robot remotely.
“While our focus is currently on providing objective structure and feedback to practicing surgeons, this work could be used to help train the machine to perform micro-surgery (such knot tying, suturing) autonomously at the highest level of quality,” Hendricks noted.
We are not there yet.
“Today, it is almost impossible at this stage to have a robot that can do a whole surgery automatically. It’s a very complex problem,” Bobda said. “We still need the human to assess the situation; no robot can do that at this stage.”
In RAS, the surgeon sits at a console and operates the robot using hand controls and foot pedals. The console provides a high-definition, magnified view of the surgical site. With the console, the surgeon moves the steady robot arms, providing smaller, more precise movements at the surgical site.
RAS has been around for more than two decades. With recent technological advances, combined with further progress incorporating AI technology, there is significant potential for massive growth in the field. More than 12 million robot-assisted surgeries have been performed worldwide with Intuitive’s da Vinci system; UF Health, the university’s academic health center, uses 10 Intuitive systems.
Medical professionals and researchers contend RAS patients experience fewer surgical complications (such as surgical site infections), less pain, less blood loss, shorter hospital stays, quicker recoveries, and less scarring. For surgeons, RAS provides enhanced precision, flexibility, and control.
“What a human cannot do, we believe a robot can do, maybe much better.”
Christophe Bobda, Ph.D., Professor and associate chair for academics Electrical and Computer Engineering department
“Robotic surgery is better in terms of control compared to traditional surgery,” Hendricks said, “because as a surgeon moves, say, an inch, the robot can scale that movement down so they can be ultra precise.”
The robotic apparatus also reduces surgeon hand tremors or any involuntary movements during a procedure.
“What a human cannot do, we believe a robot can do, maybe much better – applying pressure, for instance, and then being able to get feedback on how the organ is reacting to the resistance,” Bobda said. “We’re trying to make those machines safer.”
A key element of the research is assessing the skill of the surgeon through data. “By knowing the skill of the surgeon, we should be able to perform some corrective measures,” Bobda said.
“With robotic surgery,” Hendricks added, “you’re looking at an incision that is maybe 2 inches instead of an entire cross section,” Hendricks said.
“RAS training and evaluation can be significantly taxing on surgical staff,” he added. “We’re working to alleviate this with an autonomous system for assessing skill, freeing would-be-reviewers to focus even more on their patients, while providing metrics to track skills progression over time.”
Most data points from this project are gleaned from stereoscopic cameras (left and right feed) providing live views from the robot to the surgeon operating it. For example, the cameras observe how the tool tip is moving, its position and its rotation.
Bobda said it would be beneficial to have more cameras to see more angles and develop more data. This would help gain a more precise grasp on where the tools are positioned in the rubber robot hand.
The researchers also use HiPerGator, UF’s supercomputer, to refine large deep-learning models.
UF Health purchased its first robot-surgery system nearly 20 years ago.
These days, UF surgeons perform about 1,850 robot-assisted surgeries a year at UF Health Shands Hospital. Common RAS procedures at UF Health Shands include cytoreductive, urological, gynecological, transplants, thoracic, minimally invasive laparoscopy, pulmonary and acute care, noted Robert W. Nappo, assistant vice president of Perioperative Services at UF Health Shands Hospital.
“When they came to me about surgical robotics, I said, ‘Why not? This is a cool idea. Let’s do it.’ I don’t have the funding but let us work together. You do some work here on the side. Then we can go out and look for funding,” said Bobda.
“We wish we could have two or three more Ph.D. students working on this. That would be much faster,” he added.
Bobda is currently working on a funding acquisition from the National Institutes of Health and the National Science Foundation for, among other things, an RAS simulator.
“If we can gather a large dataset of those videos and data recorded from those machines, that would be extremely beneficial. I believe it would help the broader research community that is doing similar work to us, if we are eventually able to publish it,” Bobda said.
“The largest dataset in our field on this subject is quite outdated,” he added. “While there is work from other universities – similar datasets that push the boundaries – it would be extremely impactful if it came from this small town in Florida where we have this huge hospital and great medical program.”
Read the recently published abstract, “Exploring the Limitations and Implications of the JIGSAWS Dataset for Robot-Assisted Surgery.” The abstract was written by Hendricks, Bobda, Max Panoff, Kaiwen Xiao, Zhaoqi Wang, and Shuo Wang.